If my company ever grows to be as successful as Gitlab, I’m going to much more worried about …

It's a matter of focus: Having some time to tinker and try out stuff that might just fail is great. If it consumes so much development time that the must-haves deteoriate, you let down your customers.

I think it's less innovation and more failing to understand the requirements (i.e. only implementing half of what's needed, and then moving to the next thing). E.g. adding labels to gmail, but no way to have filters, so you have to manually apply them.

This part of the article resonated with me, as I used to work as a tech lead for a very similar startup to Gitlab. We were always pushed to "ship MVP's" and "move fast and break things" by product and design, which always resulted in a half-broken features getting pushed to production and left there once the sprint was over. This also led to a constant inexorable growth of tech debt, that made me feel like we'd never be able to get things remotely stable in our corner of the product. I've since come to the conclusion that it's much better to build small, flexible, feature-complete systems that can be used for multiple iterations of ideas.

If the failed ideas aren't then purged, then they become problems. And maybe they cause problems when implementing them too.